Analysis of actual observations of response processes is a routine procedure in applied econometrics but methods of dealing with purely subjective probability distributions as response 'data' have seemingly not hitherto been worked out. The models and methods developed here go some way to filling this void. They are illustrated by using information from a cross-sectional study of sixty Nepalese small-scale farmers. Means and variances of subjective probability distributions for rice paddy yields under different technologies are related to controlled inputs, such as fertiliser levels, and to relevant socio-economic aspects of farmers themselves, such as technological knowledge and farm size.